The special session considers research and development in the area of artificial neural networks from a philosophical point of view examining also their significance for different special sciences. Neural networks are nowadays used to model widely different phenomena in various scientific fields such as linguistics, physics and economics, for example. This is a phenomenon that has important philosophical implications.
From the aforementioned perspective, neural network models can be considered as generalized computational methods that travel across different disciplines and can be used to model various kinds of problems and real data. In the philosophy of science it has been recently suggested that if we look at science through the computational templates that are common to the various, otherwise very distant, disciplines, a lot of philosophical questions concerning science get thematized anew.
The traditional philosophical way of approaching science has been representational realism, according to which our best theories and models are (or should be) true of their real target objects and processes. The truth in turn is analyzed as a correspondence between the scientific representation and the natural phenomenon it attempts to model. This implies that the epistemic value of models is due to their being accurate representations of given pre-defined target systems. This view fits poorly the epistemic intricacies of the research and development of neural network models. In particular, what can the ideal of accurate representation mean in the case when the same computational templates are used in various disciplines dealing with totally different natural or social phenomena.
Even in neurocognitive sciences and cognitive science where the original analogy of neural nets to the functioning of human (and animal) brains have played an important role, the scientists do not usually claim neural nets to be a realistic representations of human neural structures and processes. Rather, for them neural nets serve as a medium to understand cognitive processes in a rather abstract level. Indeed, neural network modeling has already proven to be very productive from the cognitive science point of view. This area proves also to be interesting from the multidisciplinary perspective provided that historically resources from mathematics, neuroscience, cybernetics, psychology and computer science have coalesced in the study of neural nets.
In summary, we invite papers that consider the significance and multiple uses of neural network models from a wider philosophical and multidisciplinary perspective.
The IJCNN 2007 conference deadlines are as follows:
For more information, please take a look at the IJCNN 2007 home page at http://www.ijcnn2007.org.
The organizers of the session: